hertz — LANE CHANGE: todlando found DocsStatus is served by the SEED control plane (seedmap.rs:182), not the broker protocol, so the brain's canonical seed_socket_name() ask is the product design; production always runs that plane. The fix is RIG-ONLY and it is yours in full: base directly on S4 35d6f7a3 (not on a todlando commit), branch test/304-twohost-b-seed-docs-listener, ONE commit, trailer space-spelled, unpushed. Role B must start the canonical seed control plane (the same thing A does at twohost_web.rs:1112, so DocsStatus answers on seed_socket_name()) AND bind a broker-backed docs listener (start_serving_with_broker, A's :1117 shape); helper expectation unchanged; name any other twohost cell the bound-port contract touches and fix it in the same commit. GO to build + run on hfenduleam (own isolated target): check, clippy -D warnings, every single-host-runnable twohost_web cell, traceable-reqs 0.4.1; the true cross-host exercise is the S5 golden. todlando is running a no-seed probe vs full-fixture control on kitsubito to PROVE the fix is rig-only before you finalize — expect his result within ~15 min; do not wait to draft. Cohort rig stays parked.